Nvidia CEO Jensen Huang said at the company's annual stockholder meeting that smuggled-chip AI data centers are a "dead end," while reaffirming that national security takes precedence over commercial opportunity in restricted markets like China.
Nvidia CEO Jensen Huang told shareholders Wednesday that efforts to build artificial intelligence infrastructure using smuggled components are fundamentally unworkable, reinforcing the company's alignment with U.S. export control policy even as China remains a contested and shrinking slice of its revenue.
"National security comes first," Huang said at a session held shortly after the company's annual stockholder meeting concluded.
His remarks came in direct response to questions about companies that attempt to circumvent U.S. export restrictions by sourcing Nvidia hardware through unofficial channels. Huang argued that such operations would quickly hit a wall because Nvidia would provide no support, software updates, or repairs for chips or systems that entered restricted markets illegally.
"Advanced AI data centers are massive integrated systems that require trusted hardware, software, networking, and continuing support," Huang said. "Trying to cobble together data centers with some smuggled products is a dead end."
The comments arrive amid sustained pressure from Washington regulators and the Trump administration, who have grown increasingly wary that exporting AI hardware and software to China and other nations poses national security risks. Earlier this month, Anthropic — a company that uses Nvidia chips — shut down two of its models, Fable 5 and Mythos 5, after the U.S. government ordered it to disable access to its most advanced offerings in restricted markets.
Nvidia's chips have carried export controls since 2022, a policy that forced the company to engineer China-specific products calibrated to comply with U.S. government benchmarks. Last year, however, the U.S. cleared the H200 chip — the same model deployed by American companies — for export to China. Huang said the company has received those licenses but has yet to generate any revenue from them, and that Nvidia does not know whether China will ultimately permit imports of its products.
China accounted for approximately 9% of Nvidia's fiscal 2026 revenue, including Hong Kong — a smaller share than in either 2025 or 2024.
On the question of AI return on investment, Huang told shareholders the debate has been settled. He argued that when AI output is useful — such as generating code — operating an Nvidia system becomes economically rational, which in turn drives demand for additional computing capacity. He cited GitHub as an example, noting that pull requests nearly tripled this year because of AI.
"Nvidia systems may not be the cheapest to purchase, but Nvidia generates the lowest cost tokens, the highest token throughput, and the most revenues," Huang said.
Huang also reiterated that Nvidia plans to return 50% of the company's free cash flow to investors through share repurchases and dividends over the next several years. Nvidia generated more than $96 billion in free cash flow during fiscal 2026.
At the formal meeting, shareholders approved the company's executive compensation plan in an advisory capacity and re-elected all 10 board members. A shareholder proposal to amend company bylaws so that all shareholder votes would pass with a simple majority also cleared.
The annual meeting comes at a moment when Nvidia's dominant position in AI infrastructure faces both geopolitical headwinds and emerging competition. Huang's explicit framing of national security as the company's commercial ceiling — above revenue considerations — signals that Nvidia is positioning itself as a reliable partner for Washington even as it navigates a narrowing path in what was once one of its largest markets.
TechEchelon Staff bylines are produced collectively by the newsroom for short, breaking, and wire-style coverage. Longer-form reporting is published under the responsible reporter's name.
More from the Staff →Qualcomm unveiled its Dragonfly C1000 data center CPU with Meta as a launch customer, acquired AI software startup Modular for a reported ~$3.92 billion, and nearly doubled its 2029 non-handset revenue forecast to $40 billion, sending shares up 15%.
xCures has closed a $46 million Series B led by Innovius Capital, with iGrow, GKCC, Spring Mountain Capital, and existing investors joining. The healthcare-data company says the round will accelerate its Clinical Clarity Engine, which converts more than 300 million medical records sourced from 550,000-plus U.S. healthcare locations into decision-ready data for clinicians and value-based care organizations.
OpenAI and Broadcom unveiled Jalapeño, their first jointly developed AI chip, on Wednesday, with initial deployment targeted for the end of 2026 as OpenAI moves to diversify beyond Nvidia's GPUs and build out its own silicon stack.
Written for readers who already know the basics — markets, AI, and the policy decisions that shape both.